Every few weeks someone asks the same question: "I have a cheap VPS — how do I make money with it?" The honest answer, after actually doing it: the VPS is the least interesting part. Here's the full inventory of what I put on a 2 vCPU / 2 GB server, what it cost, and where the money actually is.
What the server is not
Before building anything I measured the classic "passive income" plays on a 2 vCPU box. Mining: no GPU, a dollar or two a month at best, and most providers ban it. Bandwidth-sharing: pennies on datacenter IPs, plus terms-of-service risk. Proxy resale: banned, and buyers churn the moment their IP is blocked.
Those are not businesses. They're ways to risk an account for coffee money.
What I actually deployed
The server runs the boring infrastructure of a one-person services business:
| Layer | What it is | |---|---| | Web server | Caddy with automatic HTTPS (Let's Encrypt, TLS 1.3, HTTP/2), security headers, gzip | | Pages | Bilingual landing page, 3 industry pages, ROI calculator, 3 sample-deliverable pages, 2 articles, privacy page | | API | ~500 lines of zero-dependency Python: lead capture, validation, dedupe-friendly storage, per-IP rate limits, honeypot, admin token | | Data | SQLite (WAL) — one file, backed up daily, 14-day retention | | Ops | systemd units, log rotation, daily cron backup, one-command probe, deploy script | | Internal tools | Admin dashboard (leads, statuses, CSV export, traffic stats), CLI, workflow QA checker, PDF quote/invoice generators |
No Kubernetes. No Redis. No Docker. It idles at roughly 30 MB of RAM for Caddy plus 12 MB for the API, on a box that also has swap. The whole thing would run on a $5 plan; the 2 GB matters only when you start hosting client work.
What it cost
- Money: a low-end 2 vCPU VPS in the $10–20/month class. That is the entire cash cost. Domain is
optional (I used a free sslip.io hostname for HTTPS while testing; a real domain is ~$10/year).
- Time: about two days of building, most of it spent on the parts nobody sees — error handling,
backups, the acceptance checklist, the deploy script. The HTML took an afternoon.
That's it. The "infrastructure" line item is trivially small. Which is exactly why it can't be the business.
The uncomfortable part: zero visitors
Here is the dashboard after deployment:
views: 0 visitors: 0 leads: 0 conversion: 0.0%A perfectly built site with no traffic is a business card in a drawer. Infrastructure does not create demand. Nobody wakes up wanting "fixed-scope workflow automation"; they want their invoice entry to stop eating an hour a day, and they find out you exist because you showed up in their inbox, their subreddit, their LinkedIn feed, or their search results.
The math that actually applies (from tracking my own outbound): roughly 50 targeted touches → 2–5 replies → 1 call → 0.2–0.5 sales. Ten working days of consistent outbound is the entry ticket. No amount of server tuning changes that ratio.
So what is the server actually for?
Three jobs, in order of value:
- Delivery. Client automations run on it (or on theirs). Hosting is a $29–49/month line item with
~85% margin. That's recurring revenue the site itself will never produce.
- Proof. A live page with real sample outputs — a lead-intake demo that shows duplicates being
merged, an invoice pipeline that stops a bad OCR scan — converts far better than a PDF.
- Measurement. Anonymous page analytics plus lead source attribution tells you whether the
LinkedIn post or the cold email produced anything. Without that, you're guessing.
Notice what's missing: "generate traffic." That's you, in public, repeatedly.
What I'd do differently
Start selling before building. I'd write the offer and the three sample deliverables first, send twenty messages, and only then build the site around whatever got replies. The site is a conversion tool, and conversion tools are cheap to build once you know what people actually want.
The second mistake would be pricing by the hour. Fixed scope with a written definition of done is what makes a $1,200 project deliverable in a week without an argument at the end.
What's live now
The stack, the sample outputs and the pricing are public here: flowsmith.43.153.86.167.nip.io. If you're a small team drowning in a specific copy-paste job, the ROI calculator tells you in ten seconds whether it's worth building — including when the honest answer is "don't."
The server didn't make me money. The server made it possible to deliver the thing that makes money. That's a smaller claim, and a much more useful one.